AI architecture.
Learning research.
Products.

• Architecture
• Multi-agent systems
• Orchestration
• Role Fabric

Coordinates people, agents, models and tools through work routing, execution, memory and review. Role Fabric defines responsibilities and handoffs within the wider Conduir architecture.

• Memory systems
• Contextual retrieval
• Provenance
• Model continuity

Persistent memory and contextual retrieval with provenance and scope controls. Connects working context with longer-lived records to support continuity across tasks, collaborators and models.

• Graph intelligence
• Code analysis
• Change impact
• Shadow graphs

Maps code, symbols and dependencies into a queryable graph. Change-impact analysis and isolated shadow graphs help agents investigate how a proposed change could affect the wider system.

• Exploration research
• Idea lineage
• Critique
• Synthesis

AI perspectives propose, challenge and combine ideas. A lineage graph preserves their development, disagreements and refinements, with a viewer and exports for further exploration.

• Neural research
• Transfer learning
• Selective training
• Retention

Experiments with reusable neural components and selective training investigate learning effort, transfer and retention—including where reuse works and where generalisation breaks down.

• Research & application
• Neural agents
• Game theory
• Evaluation

Game engines, reference solvers and neural agents support comparative decision-making research. The Matiz poker app adds odds and outs, connecting research with an experience of play.

• Creation tools
• AI-assisted building
• Game systems
• Browser experiences

Reusable templates and shared save, input and layout systems support AI-assisted game creation. Browser checks and release preparation help turn ideas into playable experiences.

• Engineering practice
• Cross-model review
• Adversarial critique
• Evaluation

Cross-vendor AI models propose findings that are challenged, deduplicated and ranked. Concerns are checked against the work rather than accepted through model agreement alone.

• Delivery practice
• Agent workflows
• Parallel delivery
• Quality gates

Coordinates AI contributors through separate implementation and review, leased workspaces and evidence for completion. A delivery practice for concurrent software engineering with human responsibility.

• Spatial research
• 3D interfaces
• Physics
• Shared worlds

Combines a browser-based 3D interface, shared scene state and physics. Signed operations and policy checks support research into attributable collaboration between people and AI.

• Learning research
• Quantum-inspired methods
• Mathematics
• Representations

Investigates quantum-inspired learning and mathematical representations beyond language models. The collection connects learning approaches with related memory research and open experimental questions.

• Research designs
• Graph neural networks
• Adversarial learning
• Co-learning

Research designs explore graph neural networks, adversarial learning and multi-agent wargaming. AGNN co-learning maps connect objectives, capabilities and the evidence needed to evaluate them.

• Research frontier
• Adaptive systems
• Emergence
• Self-organisation

Investigates adaptive organisation, model coordination and emergence. Connects measurable system behaviour with open questions about self-understanding and consciousness.

• Product concept
• Document AI
• Human review
• Source traceability

A proposed document-intelligence workflow connects source intake, model comparison, exploration, human review and delivery. The interactive journey makes each responsibility visible.